Deep Insights into Mango Scab: An AI-Powered Approach for Detection and Classification

Abstract

The mango, which comes from the plant genus Mangifera indica L., is a fruit crop that is well-regarded all over the world due to its scrumptious flavor and noteworthy nutritional profile. The illness known as mango scab, which is brought on by the fungus Elsinoe mangiferae, is a serious obstacle for the mango business. This disease causes the mango fruit to develop scaly lesions, which in turn results in decreased market value and yield losses. To effectively manage a disease, timely discovery and correct assessment of the severity of the condition are both essential. This study provides a hybrid deep learning (DL) model that combines a Convolutional Neural Network (CNN) for the detection of disease with a Support Vector Machine (SVM) for the categorization of the severity of the disease. The model was trained and tested using a broad dataset of mango photos that included a range of disease severity levels as well as different varieties of mango. The model's performance in illness detection resulted in an overall accuracy of 95.93%, with scores of best precisions of 97.17% and best recall of 96.28%, respectively. When it came to the classification of severity, the SVM classifier was able to achieve an accuracy of 95.93%. These findings provide evidence that the model is capable of precisely identifying the degree of mango scab illness and demonstrating its usefulness in detecting the condition. To provide mango farmers with a flexible instrument, the hybrid method makes use of the capability of CNNs to extract features as well as the SVM's skill in classifying data. This research provides growers with up-to-date information that enables targeted disease management. As a result, mango crop yields are increased, and economic losses are decreased.

Abstract from the paper’s OpenAlex record, shown for indexing and research. Copyright remains with the authors or the publisher. Read at the source ↗

The text the model read, with the source record’s inline markup (italics, sub- and superscripts) rendered and its character entities decoded; quote offsets refer to the raw text.